Chemprop Files

Notable changes

In [#1090], we started the process of integrating logging into the core code. This will make it easier for users to control what information Chemprop prints to output. It will also make it easier for developers to include more information outputs for potential debugging.

Scipy 1.15 subtly change how logit works which caused some of our tests to fail (as the values reported were slightly different than before). The expected test values have been updated. [#1142]

A new example notebook has been added which demonstraits how to adapt Chemprop to work with Shapley value analysis. This is another method to lend some intepretability to Chemprop models by highlighting which atom/bond features are most impactful to the final prediction value. [#938]

We continue to try to make chemprop easy to use. In [#1091] and [#1124] we added better warnings and error messages. And in [#1151] we made is easy to open the example notebooks in Google Colab. This allows people reading the docs to immediately jump in and try chemprop without needing to set up a python environment.

Bug Fixes

In [#1097], we fixed a bug where the transforms for scaling extra features/descriptors were turned off during validation. This caused models trained with these extra inputs to not report accurate metrics during training, which is a problem if the "best" model is selected instead of the last model as is done in hyperparameter optimization. Training a model and using the last model was unaffected as was doing inference.

[#1084] fixed a bug where R2Score did not have the attribute task_weights. This attribute is not used but is needed for compatability with other metrics

In v2.1 we transitioned to using torchmetrics for our metrics and loss functions, in part because it takes care of training across multiple nodes (DDP) automatically. Our custom metric for Matthew's correlation coefficient however was not set up the way torchmetrics expected. This was fixed in [#1131].

What's Changed

• splits file is json by @KnathanM in https://github.com/chemprop/chemprop/pull/1083
• add more helpful warnings about the splitting api change by @JacksonBurns in https://github.com/chemprop/chemprop/pull/1091
• Fix: Splits file can have multiple splitting schemes by @KnathanM in https://github.com/chemprop/chemprop/pull/1086
• Set all transforms to train during validation by @KnathanM in https://github.com/chemprop/chemprop/pull/1097
• updated warning to logger by @twinbrian in https://github.com/chemprop/chemprop/pull/1090
• Add task weights to r2score by @KnathanM in https://github.com/chemprop/chemprop/pull/1084
• Fix tracking_metric overwrite issue by @shihchengli in https://github.com/chemprop/chemprop/pull/1105
• Fix save_individual_predictions with ensembling by @shihchengli in https://github.com/chemprop/chemprop/pull/1110
• Add a helpful warning when invalid SMILES are passed by @JacksonBurns in https://github.com/chemprop/chemprop/pull/1124
• Fix batch size calculation for multicomponent by @KnathanM in https://github.com/chemprop/chemprop/pull/1098
• Not use transform_variance for unscaled targets by @shihchengli in https://github.com/chemprop/chemprop/pull/1108
• Add output size to attentive hparams by @KnathanM in https://github.com/chemprop/chemprop/pull/1133
• Fix test failure due to scipy logit by @KnathanM in https://github.com/chemprop/chemprop/pull/1142
• fix docs about extra atom descriptors by @KnathanM in https://github.com/chemprop/chemprop/pull/1139
• Fix MCC for DDP and multitask by @KnathanM in https://github.com/chemprop/chemprop/pull/1131
• V2: Add Shapley Value notebook for interpretability by @oscarwumit in https://github.com/chemprop/chemprop/pull/938
• add notebooks to colab and docs by @KnathanM in https://github.com/chemprop/chemprop/pull/1151

Full Changelog: https://github.com/chemprop/chemprop/compare/v2.1.0...v2.1.1